Beyond The Single Neuron Convex Barrier For Neural Network Certification

We propose a new parametric framework, called k-ReLU, for computing precise and scalable convex relaxations used to certify neural networks. The key idea is to approximate the output of multiple ReLUs in a layerjointly instead of separately. This joint relaxation captures dependencies between the inputs to different ReLUs in a layer and thus overcomes the convex barrier imposed by the single neuron triangle relaxation and its approximations. The framework is parametric in the number of k ReLUs it considers jointly and can be combined with existing verifiers in order to improve their precision. Our experimental results show that k-ReLU enables significantly more precise certification than existing state-of-the-art verifiers while maintaining scalability.